Biomedica : revista del Instituto Nacional de Salud最新文献

Introduction. Histological and molecular analysis of breast tissue is essential for the diagnosis, prognosis, and treatment of breast cancer. Key biomarkers include progesterone and estrogen receptors, as well as the human epidermal growth factor receptor 2 (HER2). HER2 overexpression indicates an aggressive subtype of breast cancer but enables targeted therapies that improve survival rates. However, its evaluation faces challenges, ranging from sample quality to interpretation variability. The College of American Pathologists classifies HER2 overexpression into four categories, but variations around the 10% expression threshold can lead to misinterpretations.Objective. To present an automated technique for classifying HER2-overexpressing cells in histological slides.Materials and methods. The Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology was applied using samples of 89 patients from the Unidad de Diagnóstico en Patología, covering all four HER2 expression levels. Deep learning techniques were employed, leveraging neural networks and vision transformer models through transfer learning. Additionally, a usability evaluation was conducted on the final version of the software.Results. The ViT-B/16 model achieved a classification accuracy of 90,65%, while the tool was evaluated with an acceptable level of satisfaction in its clinical application.Conclusion. Artificial intelligence demonstrated high accuracy and consistency in HER2 classification, reducing diagnostic variability and improving objectivity. However, further optimization of processing efficiency is required for broader applicability.

Introduction: Synthetic data in healthcare is an alternative for generating clinical records that resemble those registered in real clinical scenarios. The benefits of synthetic data are: greater volume of data, the possibility of representing specific patient populations, protection of real-data privacy, and improved data-sharing among different actors.

Objective: To formulate a synthetic data generation model for the gestational care process in Colombia and adapt it to the Observational Medical Outcomes Partnership (OMOP) common data model to facilitate its integration into artificial intelligence applications in maternal health.

Materials and methods: We conducted a case study of fully synthetic data formulation that included some of the most frequent outcomes and conditions during gestation based on a typical care process for pregnant women in Colombia. This approach was complemented by the generation of a common data model to facilitate data integration in future artificial intelligence applications or complementary systems that benefit from a standardized language, regardless of the system or form of classification.

Results: We formulated a model for the synthetic generation of clinical data –applicable to real clinical settings– that spans the entire gestational care until the perinatal period. The model included the most frequent clinical conditions and outcomes, which were diagrammed in the Synthea™ tool with their corresponding clinical probabilities of occurrence based on the reported literature or the usual practice of obstetric specialists in Colombia.

Conclusions: This study demonstrates that the generation of synthetic data applied to the gestational care process in Colombia was feasible and represents a pioneering contribution in the region

Introduction: Anemia is a frequent complication in patients with chronic kidney disease undergoing hemodialysis and is associated with increased morbidity, mortality, and healthcare burden. Accurate classification is essential to optimize treatment with intravenous iron and erythropoiesis-stimulating agents. Rule-based clinical decision support systems (CDSS) provide a strategy to standardize this process.

Objective: To describe the development and implementation of a knowledge-based clinical decision support system for the automated classification of anemia in hemodialysis patients using laboratory data.

Materials and methods: This retrospective observational study included 883 adult patients receiving prevalent hemodialysis during 2023. An algorithm was developed based on established clinical guidelines [Sociedad Latinoamericana de Nefrología e Hipertensión (SLANH)], KDIGO, NICE to classify patients with hemoglobin below 12 g/dl into three categories: absolute iron deficiency, functional iron deficiency, and candidates for therapeutic trial with intravenous iron. The system also flagged cases with suspected severe secondary hyperparathyroidism (PTH > 800 pg/ml). Data was obtained from the laboratory information system and the clinical decision support system. We applied a descriptive statistical analysis.

Results: The clinical decision support system automatically classified patients into the following categories: functional iron deficiency (39.2%), severe hyperparathyroidism (26.7%), absolute iron deficiency (17.7%), and candidates for intravenous iron trial (16.4%). A subgroup (9.5% within the functional iron deficiency group) also showed elevated PTH levels, suggesting potential resistance to erythropoiesis-stimulating agents. Distinct clinical profiles were observed across the groups.

Conclusions: The clinical decision support system enabled automated and standardized classification of anemia in hemodialysis patients, supporting evidence-based clinical decision-making. Its implementation represents a digital health innovation with the potential to improve the quality and safety of anemia management in chronic kidney disease.

Introduction: Conventional cervical cytology (Pap smear) remains a primary method for cervical cancer screening in Colombia, despite limitations in diagnostic yield and heavy workload. The potential of artificial intelligence to address these challenges is yet to be evaluated in our population.

Objective: To evaluate and compare the discriminative ability of four artificial intelligence-based models for the detection of abnormalities in Pap smears.Materials and methods. A total of 650 images of Pap smear cells were obtained from a university cohort in northeastern Colombia. These images were subjected to diagnostic evaluation by an expert pathologist. Four artificial intelligence models (DenseNet, InceptionV3, MobileNet, and VGG19) were trained using data from a publicly available Pap smear database with digital image analysis and deep learning. The discriminative ability of the models was determined by calculating their sensitivity, specificity, and area under the curve.

Materials and methods: MobileNet tuvo la mejor capacidad discriminativa [área bajo la curva (AUC) de 0,97) con una especificidad del 0,99 y sensibilidad de 0,78 para la detección de alteraciones en la citología cervicouterina. Por otro lado, InceptionV3 tuvo un mejor desempeño en el tamizaje, con sensibilidad del 0,93, especificidad de 0,82 y área bajo la curva de 0,947.

Results: MobileNet showed the highest discriminative ability (AUC = 0.97), with a specificity of 0.99 and sensitivity of 0.78 for the detection of altered cells in Pap smears. On the other hand, InceptionV3 had the best performance capabilities for screening, with a sensitivity of 0.93, specificity of 0.82, and AUC of 0.947.

Conclusions: The results of this study illustrate the advantages and disadvantages of different artificial intelligence models and how their application could help improve the diagnostic performance of manual reading in cervical cancer screening or even serve as a primary screening method to rule out negative cases, by achieving a diagnostic performance comparable to that of manual reading.

Introduction: Medical staff often face difficulties in consulting and applying clinical guidelines in practice. Large language models, especially when combined with retrieval-augmented generation, may help overcome these challenges by producing context-specific outputs with improved adherence to medical guidelines.

Objectives: To assess the performance of commercial large language models in answering maternal health questions within retrieval-augmented generation systems, using both human and automated evaluation metrics.

Material and methods: A controlled experiment was designed to obtain accurate, consistent answers from a retrieval-augmented generation system based on Colombian maternal care guidelines. A physician formulated ten questions and defined the groundtruth answers. Various large language models were tested with a standardized prompt and evaluated through binary answer-concept ranking and retrieval-augmented generation assessment, metrics, judged by two independent large language models.

Results: Generative pre-trained transformer 3.5 (GPT-3.5) achieved the highest physicianassessed accuracy (0.90). Claude 3.5 obtained the top faithfulness score (0.78) under GPT-4.o evaluation, while Mistral ranked highest (0.84) under Claude 3.5 evaluation. Regarding answer relevance, GPT-3.5 scored highest across both judges (0.94 and 0.86).

Conclusions: Integrating retrieval-augmented generation into obstetric care has the potential to enhance evidence-based practices and improve patient outcomes. However, rigorous validation of accuracy and context-specific reliability is essential before clinical deployment. The findings of this study indicate that large-scale models (e.g., GPT-3.5, Claude, Llama 70B) consistently outperform lighter models such as Llama 8B.

Introduction: PIWI-interacting RNAs are small and non-coding RNAs involved in gene regulation and transposable element repression, emerging as critical biomarkers and therapeutic targets in oncology. Advances in artificial intelligence, such as recurrent neural networks, long short-term memory networks, and graph convolutional networks, offer significant improvements in PIWI-interacting RNA detection.

Objectives: To evaluate the performance of artificial intelligence models, including recurrent neural networks, long short-term memory, and graph convolutional networks, in detecting PIWI-interacting RNAs and assessing their implications for cancer diagnostics and prognosis.

Materials and methods: A systematic review of 24 studies was conducted across PubMed, ScienceDirect, Scopus, and Web of Science, focusing on artificial intelligence-based approaches for PIWI-interacting RNA detection. Inclusion criteria were original articles published in English or Spanish using artificial intelligence models in clinical or experimental settings. Performance metrics such as accuracy, sensitivity, and specificity were analyzed.

Results: Long short-term memory models achieved the highest overall accuracy (92.3%), followed by graph convolutional networks (91.4%), support vector machines (88%), and recurrent neural networks (85.7%). Sensitivity and specificity were also highest in long short-term memory (94% and 91%, respectively). Graph convolutional networks showed superior performance in identifying PIWI-interacting RNA-disease associations with complex datasets. Support vector machine models were effective in smaller datasets but exhibited scalability limitations.

Conclusion: Artificial intelligence models, especially long short-term memory and graph convolutional networks, significantly enhance PIWI-interacting RNA detection, supporting their application in cancer diagnostics and personalized medicine. Future studies should refine these models, address dataset biases, and explore their integration into clinical workflows.

Introduction: Enteric viruses significantly impact morbidity, mortality, and healthcare. Transmission through wastewater is favoured in highly contaminated areas due to inadequate treatment.

Objective: To determine the number of rotaviruses and their infectious capacity from wastewater samples used for irrigation in the western part of Bogotá.

Materials and methods: Concentrations of group A rotavirus were monitored in wastewater using molecular methods. The infectivity of rotaviruses was evaluated in a mouse intestinal villi model. We assessed the feasibility of applying this approach for environmental health surveillance in Colombia, considering findings reported by other authors.

Results: The research focused on the La Ramada irrigation network in the western part of Bogotá, specifically the Canal San José. We analysed eighteen wastewater samples using qRT-PCR and detected group A rotavirus in twelve of them. The positive samples contained infectious rotavirus, as confirmed through the mouse villi model.

Conclusion: This study shows that contamination by group A rotavirus is frequent in wastewaters from the Canal San José in the La Ramada irrigation network in the western part of Bogotá and reveals high concentrations of rotavirus. The results suggest that villi from mouse intestines serve as a reliable model for isolating rotavirus from wastewaters. These findings provide a new approach for environmental health surveillance in Colombia, based on molecular epidemiology for waters highly contaminated with human enteric viruses.

Introduction: The Lancet Countdown 2023 report for Latin America indicates that rising temperatures influence the transmission of the dengue virus. In Colombia’s Caribbean region, a significant association has been identified between dengue incidence and climatic variables, such as temperature, humidity, and precipitation.

Objective: To analyze the relationship between the incidence of dengue and the oceanic Niño index in the departments of the Colombian Caribbean region from 2021 to 2023.

Materials and methods: An ecological time series study was conducted using distributed lag non-linear models and autoregressive integrated moving average models in the seven departments of the Caribbean region. Descriptive and autoregressive analyses were performed using JASP and RStudio. Non-linear and lagged analyses were run with the dlnmpackage in RStudio.

Results: A positive and significant relationship between the oceanic Niño index and dengue incidence was found for 2023 data, the year when the El Niño - ENSO (El Niño-Southern Oscillation) warm phase occurred. Bolívar, Cesar, Córdoba, and Magdalena departments showed positive correlations. A non-linear relationship between El Niño/La Niña and dengue incidence was also observed, with a higher increase in dengue cases during El Niño events.

Conclusions: The oceanic Niño index appears to be a useful climatic indicator for monitoring increases in the monthly dengue incidence rate in the analyzed departments of Colombia’s Caribbean región.